Tuning a digital multivariable controller for a lab-scale helicopter system via simulated annealing and evolutionary algorithms

Abstract

Helicopter systems are considered a complex and challenging control problem due to strong couplings and high non-linearities. In this paper, simulated annealing (SA), as one of the leading methods in search and optimization, is applied to tune a multivariable controller of a lab-scale helicopter system. The lab-scale helicopter system is a multivariable experimental aerodynamic test rig that resembles the behaviour of a real helicopter. The control objectives are quickly to reach a desired position or track a trajectory. A centralized cross-coupled PID controller is used to achieve these objectives. First, SA optimizations are carried out with 24 different initial configurations. Then, the best results of these SA configurations are compared with other controllers obtained with evolutionary algorithms (EAs) of genetic algorithms (GAs), modified particle swarm optimization (MPSO) and differential evolution (DE). The comparisons are based on statistical measures of 20 independent trials, non-linear computer simulations of different input signals and real-time measurements for various commands of positions or trajectories. Results show that SA obtained the best performance index and acceptable time-domain performance on reaching hovering point, following a step command and tracking a sine trajectory compared with the investigated EAs.